The Pros and Cons of Smoothly Scaling AI Capabilities
RyanGreenblatt · x · 2026-07-11
He argues that the advantage of a "Plan A" is that capability improvements remain steady rather than spiking, providing an opportunity to discover and fix issues at each stage. Another benefit is that humans can research and iterate at a given capability level for years before truly delegating decisions to AI.
However, he stresses that this doesn't guarantee safety: even after years of researching AI alignment, intelligence, and epistemological properties, relying on AI could still be highly dangerous. A faster capability leap would only exacerbate the situation.
More from AGI Musings
- AI era exposes how often human pride beats basic reasoning — haider1 · 2026-07-21
- FactoryAI’s Enoreyes says model distillation is basically unstoppable — LangChain · 2026-07-21
- Andrew Blumberg says formalization without interpretability is not science — AlexKontorovich · 2026-07-21
- Ken Ono says AI is forcing mathematicians to rethink how discovery works — soumitrashukla9 · 2026-07-21
- Open-source labs could distill a state-of-the-art model to 32GB or 80GB VRAM, the post argues — bookwormengr · 2026-07-21
- Two US companies are now using superintelligence to speed up the next generation of models — yacineMTB · 2026-07-21